case studies rman metrics · tuning the large pool for rman • carefully identify relationships...

LI SIG Fall 2008 Meeting

Anthony D Noriega, MSCS, MBA, OCPAdvanced Data Networks

Case Studies in

RMAN Metrics

Long Island SIG (Fall 2008)

Speaker Qualifications

• Independent Consultant, ADN• Speaker at NYOUG meetings• 24 years of IT experience• 18 years of Oracle experience, 13 as a DBA • RMAN experience with Oracle8i,9i, 10g, and

11g, since 1999. • MS Computer Science, NJIT, 1993• PhD CIS candidate, NJIT, 1997• MBA MIS, Montclair State University, 2006


Objectives

• Present an proven approach to RMAN time and storage space metrics that guarantees robustness in backup/recovery, disaster recovery (DR), and business continuity (BC) processes.

• Derive a series of technical arguments to support an DR/BC optimization methodology.


Best Practices• Determine the backup window of time and

related duration Establish the restore window of time and related duration

• Establish a process-related cost analysis• Create a policy framework in order to

consistently maintain archived redo log files in backup sets, or available accordingly in the file system.

• Discuss a tuning model for the large pool.


Best Practices

• Utilize available features and capabilities such as Block Change Tracking, control file auto-backup, native backup compression, or point-in-time recovery when appropriate.

• Establish DR/BC a historic log.


Experimenting with RMAN

• Production environments– With recovery Catalog

• Hitachi SAN• MPP (128-,32-processor logically

partitioned Solaris servers ).• Maximum Availability Architecture

(MAA) using Hardware Assisted Resilient Data (RAC plus Data Guard emulation).


Experimenting with RMAN

• Production environments– Without recovery Catalog

• Undisclosed SAN architecture• MPP (16 and 32-processor Solaris

servers, but comparable cohort architecture).

• Oracle Data Guard in place.


Tuning with Large Pool RMAN

• Production environments– Linux 4-node RACs, 9i and 10g

• Undisclosed SAN architecture• MPP (Linux Intel 64-bit 8/16 dual-core

processor RH Linux servers).• Oracle RAC and Data Guard in place

(MAA).


Architecture Environments

• Oracle9i• Oracle10g experimental projections• fast_start_mttr_target custom settings• Block Change Tracking (Oracle10g/11g)• Incremental backup level 0 mostly.


The Shell

RMAN Backup shell for Oracle Instances Studied (Without a Recovery Catalog)


Backup Size and Duration

• RMAN Backup Duration for Oracle Instances Studied (With a Recovery Catalog)


RMAN Backup Size and Duration

• RMAN Backup Duration for Oracle Instances Studied (without a Recovery Catalog)


RMAN Backup Size and Duration

• RMAN Backup Duration for Oracle Instances Studied (NOCATALOG expanded view)


Backup Size Duration Line Chart


NOCATALOG Restore Metrics

• RMAN Restore Duration for Oracle Instances Studied (Without a Recovery Catalog)


The Shell

• RMAN Backup shell for Oracle Instances Studied (With a Recovery Catalog)



• RMAN Backup Duration for Oracle Instances Studied (With a Recovery Catalog)



• RMAN Backup Duration for Oracle Instances Studied (With Backup Validation Using the Recovery Catalog)



• RMAN Restore Duration for Oracle Instances Studied (With Backup Validation Using the Recovery Catalog)


Data Collected for Analysis


A Simple RMAN Regression Model


Further Model Regression


Correlation (Variable Interaction)


Oracle10g Improved Adjustments• The number of blocks changed as a

percent of the total backup (incremental policies).

• Global block density being backed up.• The Global block density in the database

being backed up.• The combined effect of all previous

factors.• The settings of block change tracking.


Oracle10g Improved Adjustments• Settings for:

– Fast_start_mttr_target– Log_checkpoint_interval– Log_checkpoint_timeout– Large_pool_size

• V$MTTR_TARGET_ADVICE view


Some Oracle10g RMAN ObjectsOBJECT_TYPE OBJECT_NAME----------------- -----------------------------------PACKAGE CURSORMANAGERINTERFACEPACKAGE GENCURSORMANAGERINTERFACETYPE GENCURSORMANAGERSEQUENCEVIEW GV_$RMAN_CONFIGURATIONVIEW GV_$RMAN_OUTPUTVIEW V_$RMAN_BACKUP_JOB_DETAILSVIEW V_$RMAN_BACKUP_SUBJOB_DETAILSVIEW V_$RMAN_BACKUP_TYPEVIEW V_$RMAN_CONFIGURATIONVIEW V_$RMAN_OUTPUTVIEW V_$RMAN_STATUS


RMAN NOCATALOG Configuration


RMAN Session Long Operations


Listing the Backup Contents


DBPITR Tree Recovery Path


Model Implications on VLDBs

• Model Variable Correlation• Support for HARD, SAN-based, Dataguard,

and RAC solutions.• Proven metric correlation in Maximum

Availability Architecture (MAA).


Business Analysis Considerations

• Disaster Recovery (DR)• Business Continuity (BC)• Backup and Recovery (BR)• SDLC • Overall Preparedness for any RMAN

process.


Business Analysis Considerations• Quality Assurance

– Backupset approach– Archived redo log backup policy– Retention policy strategy.

• Shell Robustness throughout OS platform• Shell scheduling, deployment with

periodic validation and verification.


Tuning the Large Pool for RMAN• Carefully identify relationships

between the large pool size setting and the RMAN operation (backup/restore) duration, and the backup pieces size and the overall backup size.

• Large Pool Size’s Oracle recommended setting is 16M.

• Findings are preserve standalone or using third-party tools or utilities.


Tuning the Large Pool for RMAN


Concluding Remarks

• Correlation between database size/backup size and duration.

• Factors:– Environment– Backup Validity

• Preparedness.


Concluding Remarks

• DBPITR• Backup Restore/Recovery Time• Reliability=MTTF/(MTTF+MTTR)• Business Process Robustness


Concluding Remarks

• Oracle9i and Oracle10g congruent Metrics• Configuration Bottlenecks• Planning and Risk Analysis• Block Change Tracking• BR/DR/BC Business Process.• The large pool size is directly proportional

to the backup size and inversely proportional to the backup duration.


Concluding Remarks


Questions?


Thank You• Please complete your evaluation form!• Speaker: Anthony D. Noriega• Title: Case Studies in RMAN Metrics• Special Session• Contact information: • [email protected]

case studies rman metrics · tuning the large pool for rman • carefully identify relationships...

Documents